Flash Storage Passthrough for Virtual Machines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtualization systems, the use of virtual disks for accessing memory devices introduces processing overhead and latency due to context switching and indirect access methods, which can restrict access to memory devices for multiple virtual machines.
Innovation Solution
Implementing flash storage device passthrough by allocating queue resources of a memory device, such as a UFS device, to respective virtual machines, allowing direct access and control, while ensuring isolation among queues to prevent conflicts and reserving management functions for the hypervisor.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If virtual disks are used for accessing memory devices, then multiple virtual machines can access storage resources, but processing overhead and latency increase due to context switching and indirect access methods
Solution Approach 1:
The patent segments the queue resources of the flash storage device into multiple independent queues, with each queue dedicated to a specific virtual machine. This segmentation eliminates the need for context switching between VMs while maintaining direct hardware access, thereby reducing latency while preserving multi-VM adaptability
2Loss of time
If direct device passthrough is implemented, then access latency is reduced, but resource isolation and management control are compromised
Solution Approach 1:
The patent segments the flash storage device's queue resources into multiple independent queues, each assigned to a specific virtual machine. This segmentation provides both direct access (reducing latency) and resource isolation (simplifying management), as each VM has dedicated queues that cannot interfere with others
Solution Approach 2:
The patent introduces a new dimension of queue resource allocation, moving from a single shared queue model to a multi-queue model where queues exist in a separate resource dimension. This allows simultaneous direct access for multiple VMs while maintaining isolation through the queue dimension boundary
3Productivity
If multiple virtual machines share the same queue resources, then resource utilization is improved, but access conflicts and performance degradation occur
Solution Approach 1:
The patent segments the queue resources into dedicated queues for each virtual machine, eliminating access conflicts while maintaining high resource utilization. Each VM has exclusive access to its assigned queues, ensuring stable and predictable performance without degradation from sharing
Data Source
AI summary
In some implementations, a memory system may configure an association of queue resources to identifiers of respective virtual machines of one or more virtual machines, wherein the queue resources are associated with multiple queues of a universal flash storage (UFS) host. The memory system may receive a request indicating an identifier and one or more queue resources. The memory system may perform an action, via a UFS device, associated with the request based on the identifier, the one or more queue resources, and the association.